4 require 'regen/regen_lib.pl';
5 require 'regen/charset_translations.pl';
7 # Generates the EBCDIC translation tables that were formerly hard-coded into
10 my $out_fh = open_new('ebcdic_tables.h', '>',
11 {style => '*', by => $0, });
13 sub output_table ($$) {
14 my $table_ref = shift;
17 # Tables in hex easier to debug, but don't fit into 80 columns
20 die "Requres 256 entries in table $name, got @$table_ref" if @$table_ref != 256;
22 print $out_fh "EXTCONST U8 $name\[\] = {\n";
24 print $out_fh "/* _0 _1 _2 _3 _4 _5 _6 _7 _8 _9 _A _B _C _D _E _F */\n" if $print_in_hex;
25 for my $i (0 .. 255) {
27 printf $out_fh "/* %X_ */ ", $i / 16 if $i % 16 == 0;
28 printf $out_fh " 0x%02X", $table_ref->[$i];
31 printf $out_fh "%4d", $table_ref->[$i];
33 printf $out_fh " /* %X_ */", $i / 16 if $print_in_hex && $i % 16 == 15;
34 print $out_fh ",", if $i < 255;
35 print $out_fh "\n" if $i % 16 == 15;
37 print $out_fh "/* _0 _1 _2 _3 _4 _5 _6 _7 _8 _9 _A _B _C _D _E _F */\n" if $print_in_hex;
38 print $out_fh "};\n\n";
43 #ifndef H_EBCDIC_TABLES /* Guard against nested #includes */
44 #define H_EBCDIC_TABLES 1
46 /* This file contains definitions for various tables used in EBCDIC handling.
47 * More info is in utfebcdic.h */
50 my @charsets = get_supported_code_pages();
51 shift @charsets; # ASCII is the 0th, and we don't deal with that here.
52 foreach my $charset (@charsets) {
53 # we process the whole array several times, make a copy
54 my @a2e = @{get_a2n($charset)};
56 print $out_fh "\n" . get_conditional_compile_line_start($charset);
59 print $out_fh "/* Index is ASCII platform code point; value is $charset equivalent */\n";
60 output_table(\@a2e, "PL_a2e");
62 { # Construct the inverse
64 for my $i (0 .. 255) {
67 print $out_fh "/* Index is $charset code point; value is ASCII platform equivalent */\n";
68 output_table(\@e2a, "PL_e2a");
71 my @i82utf = @{get_I8_2_utf($charset)};
73 /* (Confusingly named) Index is $charset I8 byte; value is
74 * $charset UTF-EBCDIC equivalent */
76 output_table(\@i82utf, "PL_utf2e");
78 { #Construct the inverse
80 for my $i (0 .. 255) {
81 $utf2i8[$i82utf[$i]] = $i;
84 /* (Confusingly named) Index is $charset UTF-EBCDIC byte; value is
85 * $charset I8 equivalent */
87 output_table(\@utf2i8, "PL_e2utf");
93 # These are invariants or continuation bytes.
94 for my $i (0 .. 0xBF) {
95 $utf8skip[$i82utf[$i]] = 1;
98 # These are start bytes; The skip is the number of consecutive highest
99 # order 1-bits (up to 7)
100 for my $i (0xC0 .. 255) {
102 if (($i & 0b11111110) == 0b11111110) {
105 elsif (($i & 0b11111100) == 0b11111100) {
108 elsif (($i & 0b11111000) == 0b11111000) {
111 elsif (($i & 0b11110000) == 0b11110000) {
114 elsif (($i & 0b11100000) == 0b11100000) {
117 elsif (($i & 0b11000000) == 0b11000000) {
121 die "Something wrong for UTF8SKIP calculation for $i";
123 $utf8skip[$i82utf[$i]] = $count;
127 /* Index is $charset UTF-EBCDIC byte; value is UTF8SKIP for start bytes;
128 * 1 for continuation. Adapted from the shadow flags table in tr16. The
129 * entries marked 9 in tr16 are continuation bytes and are marked as length 1
130 * here so that we can recover. */
132 output_table(\@utf8skip, "PL_utf8skip");
135 use feature 'unicode_strings';
139 for my $i (0 .. 255) {
140 $lc[$a2e[$i]] = $a2e[ord lc chr $i];
142 print $out_fh "/* Index is $charset code point; value is its lowercase equivalent */\n";
143 output_table(\@lc, "PL_latin1_lc");
148 for my $i (0 .. 255) {
150 if (length $uc > 1 || ord $uc > 255) {
151 $uc = "\N{LATIN SMALL LETTER Y WITH DIAERESIS}";
153 $uc[$a2e[$i]] = $a2e[ord $uc];
156 /* Index is $charset code point; value is its uppercase equivalent.
157 * The 'mod' in the name means that codepoints whose uppercase is above 255 or
158 * longer than 1 character map to LATIN SMALL LETTER Y WITH DIARESIS */
160 output_table(\@uc, "PL_mod_latin1_uc");
165 for my $i (0 .. 255) { # Initialise to identity map
166 $ascii_fold[$i] = $i;
169 # Overwrite the entries that aren't identity
170 for my $chr ('A' .. 'Z') {
171 $ascii_fold[$a2e[ord $chr]] = $a2e[ord lc $chr];
173 for my $chr ('a' .. 'z') {
174 $ascii_fold[$a2e[ord $chr]] = $a2e[ord uc $chr];
177 /* Index is $charset code point; For A-Z, value is a-z; for a-z, value
178 * is A-Z; all other code points map to themselves */
180 output_table(\@ascii_fold, "PL_fold");
185 for my $i (0 .. 255) {
189 # lc and uc adequately proxy for fold-case pairs in this 0-255
192 $uc = $char if length $uc > 1 || ord $uc > 255;
194 $latin1_fold[$a2e[$i]] = $a2e[ord $lc];
196 elsif ($uc ne $char) {
197 $latin1_fold[$a2e[$i]] = $a2e[ord $uc];
200 $latin1_fold[$a2e[$i]] = $a2e[$i];
204 /* Index is $charset code point; value is its other fold-pair equivalent
205 * (A => a; a => A, etc) in the 0-255 range. If no such equivalent, value is
206 * the code point itself */
208 output_table(\@latin1_fold, "PL_fold_latin1");
211 print $out_fh get_conditional_compile_line_end();
214 print $out_fh "\n#endif /* H_EBCDIC_TABLES */\n";
216 read_only_bottom_close_and_rename($out_fh);